The Development of AERMOD-Ready Meteorological Data for ......AERMOD over the entire three year...
Transcript of The Development of AERMOD-Ready Meteorological Data for ......AERMOD over the entire three year...
The Development of AERMOD-Ready Meteorological Data for the South Coast
Air Basin and the Coachella Valley
Final Report
Volume I
Prepared by
EnviroComp Consulting, Inc.
2298 Ocaso Camino Fremont, CA 94539
http://www.envirocomp.com
For
South Coast Air Quality Management District 21865 Copley Drive
Diamond Bar, CA 91765-4178 (909) 396-3520
April 17, 2009
Project: EC-07-015 Report: 09-04-17 (I)
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table of Contents
1 Introduction............................................................................................................................. 3
2 Summary of Deliverables ....................................................................................................... 6
3 AERMET Formulation ......................................................................................................... 15
4 Methodology......................................................................................................................... 29
5 Results................................................................................................................................... 54
6 Conclusion ............................................................................................................................ 75
2
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
1 Introduction
This is the final report submitted by Envirocomp Consulting, Inc.1 to the South Coast Air
Quality Management Division (AQMD) in partial fulfillment of the project entitled “The
Development of AERMOD-Ready Meteorological Data for the South Coast Air Basin and the
Coachella Valley”2. It describes the development of the meteorological input files required to
run AERMOD3, which replaced ISCST34 in November 2005 as the EPA recommended
regulatory dispersion model for short range applications. The AERMOD meteorological input
files, provided to AQMD on a CD accompanying this report, have been developed for 26
different regions (“sub-areas”) that cover the South Coast Air Basin and Coachella Valley. They
will fill the same role for AERMOD as those currently maintained by AQMD for use with
ISCST3, to be downloaded by modelers for regulatory applications of AERMOD within the
AQMD.
The meteorological input files for AERMOD contain hourly records of meteorological variables
needed to run the model. Regulatory applications generally require several years of data, with
five years being the standard. The hourly records include routinely measured variables such as
wind speed, wind direction and temperature, as well as variables derived from these
measurements used to characterize the mean and turbulent structure of the atmospheric boundary
layer. Examples of these derived variables are the surface friction velocity and the convective
velocity scale. The meteorological input files are the outputs of the AERMET5 processor,
developed by EPA.
The development of AERMOD input files consisted of the following steps:
1. Examine meteorological data collected throughout the Southern California area for
availability and completeness.
1 http://www.envirocomp.com2 Project Number #P2008-10 3 Cimorelli, A. J. and coauthors, 2005: “AERMOD : A dispersion model for industrial source applications. Part I:
General model formulation and boundary layer characterization”, J. Applied. Meteorol., 44, 682-693. 4 http://www.epa.gov/scram001/dispersion_alt.htm#isc3 5 http://www.epa.gov/scram001/metobsdata_procaccprogs.htm#aermet
3
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
2. Check the quality of the data, fill in gaps, and create input files for AERMET, one for
each of the 26 AQMD sub-areas within the district.
3. Run the data through AERMET to generate the surface (.sfc) and profile (.pfl) AERMOD
inputs for AERMOD.
4. Compare AERMET output with available measurements of micrometeorological
variables.
5. Process the input files using software developed by Envirocomp to further check the
hourly entries in the *.sfc and *.pfl files to flag and correct any suspicious values.
A flowchart of this procedure is shown in Figure 1.1.
As part of this project, we also developed and provided to AQMD several post-processors that
operate on the *.sfc and .pfl files output by AERMET. These post-processors can be run at the
discretion of AQMD, and perform the following functions:
1. Adjust the mixed layer height and the associated convective velocity scale in the input
files to account for cases when the one-dimensional boundary layer model in AERMET
cannot account for the two-dimensional effects seen in coastal areas, such as Long Beach.
2. Adjust AERMOD input files to account for change in micrometeorological variables
when the AERMOD application site has different surface parameters than those at the
site at which meteorological observations were made.
3. Construct input files corresponding to specified values of surface roughness, albedo, and
Bowen ratio through adjustments of the existing input files. This processor uses a method
that provides a close approximation to the files that would be produced by re-running
AERMET with the specified values of surface parameters. The primary function of this
processor is to provide a quick assessment of the effects of changes in surface parameters
on AERMOD inputs.
In the following sections of the report, we provide details of the steps used to develop the
AERMOD input files, and describe the post-processors. We also provide a brief description of
the method used in AERMET to construct AERMOD inputs.
4
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 1.1: Flowchart of basic procedure to create AERMOD input files for each sub-area within the AQMD.
5
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
2 Summary of Deliverables
The deliverables for the project are:
• AERMOD input files for each AQMD sub-area (*.sfc and *.pfl). Two sets of files are
provided. The first set assumes values for land surface parameters (surface roughness
length, Bowen ratio, surface albedo) determined by the EPA AERSURFACE processor
for each sub-area. The second set assumes constant values of these three land-surface
parameters, which are applied to all sub-areas.
• The codes for the three post-processors to check and/or modify the AERMOD input files.
• AERMET input files for each sub-area. Among these input files are the “pre-processed”
meteorological data files, which contain the meteorological inputs to AERMET.
• Raw meteorological data files, as well as various interim files and codes, to create the
pre-processed meteorological data files for each sub-area that are input to AERMET.
• Wind roses of the hourly winds used to construct the input files for each sub-area
(Appendix A, and provided separately in electronic files on the accompanying CD)
• User’s Guide (Appendix C)
The AQMD sub-areas and district surface meteorological stations that provided the wind data are
shown in Figure 2.1.
The *.sfc and *.pfl files for each sub-area are the main deliverable. As an example of these, the
leading hourly records for the Anaheim sub-area files (anah.sfc and anah.pfl) are shown in
Figure 2-2 and Figure 2-3, respectively. The full files cover the period 2005 – 2007. An
explanation of the entries in the hourly records contained in these files is as follows, taken from
the Lakes Environmental AERMOD-View6 software help-page:
6 http://www.lakes-environmental.com/
6
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
*.sfc file (see Figure 2.2)
7
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
*.pfl file (see Figure 2.3)
It is seen in Figure 2.3 that the last two columns of the anah.pfl file, which provide the hourly
values for the standard deviations of wind direction and vertical wind speed fluctuations, are
flagged as missing (99.0). This is the case for all of the *.pfl files we created. This column has
non-missing values only if measurements of these variables are read into AERMET. In absence
of these measurements, AERMOD computes these turbulence variables using methods described
in Section 3.
8
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ The period averaged concentration results obtained by running the anah.sfc and anah.pfl files in
AERMOD over the entire three year period 2005 – 2007 of hourly meteorology contained in
these files are shown in Figure 2.4 and Figure 2.5, respectively, for an elevated point source and
a surface area source. A unit emission rate of 1 g/sec was assumed in each case. The purpose of
this run was to confirm that the anah.sfc and anah.pfl files successfully port and run in
AERMOD. The *.sfc and *.pfl files developed for the other sub-areas were also tested in this
manner to be sure that they successfully port and run.
9
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 2.1: Map of Southern California, showing the geographical boundaries of each AQMD sub-area and the
location of district surface meteorological stations in the area.
10
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 2.2: Leading hourly records for the Anaheim (“ANAH”) sub-area *.sfc file output by AERMET.
11
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 2.3: Leading hourly records for the Anaheim (“ANAH”) sub-area *.pfl file output by AERMET.
12
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 2.4: Period average concentration field (μg/m3) resulting from running the Anaheim sub-area input files
(anah.sfc and anah.pfl) in AERMOD. Run assumes a unit emission rate (1 g/sec) from a point source.
13
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 2.5: Period average concentration field (μg/m3) resulting from running the Anaheim sub-area input files
(anah.sfc and anah.pfl) in AERMOD. Run assumes a unit emission rate (1 g/sec) from an area source.
14
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
3 AERMET Formulation
AERMET7 processes the raw meteorological observations (described later) to produce the
meteorological inputs for AERMOD in two files: a surface file (.sfc) and a profile file (.pfl). The
surface file contains the following variables as a function of hour of day:
1. Sensible heat flux, Hs
2. Surface friction velocity, u*
3. Convective velocity scale, w*
4. Temperature gradient above the mixed layer, γ
5. Convective boundary layer height, zic
6. Mechanical boundary layer height, zim
7. Monin-Obukhov (M-O) length, L
8. Surface roughness length, z0
9. Bowen ratio, Bo
10. Surface albedo, α
11. Wind speed
12. Wind direction
13. Temperature
The profile file contains information on the following variables as a function of height:
1. Standard deviation of the vertical velocity fluctuations, σw
2. Standard deviation of horizontal wind direction fluctuations, σθ
These variables are estimated using a one-dimensional boundary layer model that assumes
horizontally homogeneous conditions. The following sections describe relevant details of the
methods used in the model to construct the input variables.
7 http://www.epa.gov/scram001/metobsdata_procaccprogs.htm#aermet
15
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ Surface Energy Balance
AERMET is based on a one-dimensional boundary layer model, and assumes that Monin
Obukhov (MO) similarity holds near the surface. The boundary-layer is driven by the surface
energy balance given by
GHHRTTS LSNoi ++==−+− )1( α , (1)
where S is the incoming solar radiation, Ti is the incoming thermal radiation, and To is the
outgoing thermal radiation. The right hand side contains the sensible (Hs), latent (HL) and
ground (G) heat fluxes, respectively. In equation (1), α is the surface albedo, which is the
fraction of the incoming solar radiation that is reflected. So (1-α) is the fraction of the solar
radiation that is absorbed by the ground.
The sensible heat flux, Hs, plays a major role in the production and destruction of turbulence: it
determines the level of turbulence both during the day and the night, and governs the evolution
of the daytime boundary layer. It is the energy flux transferred from or to the ground, and is
estimated using the surface energy balance, which we now review.
The radiative input to the surface consists of two components: solar radiation, S, and thermal
radiation, T. Solar radiation refers to the wavelength region corresponding to the radiation of
from the sun, whose effective blackbody temperature is close to 6000°K. Most of the solar
energy lies in the wavelength region 0 ≤ λ ≤ 4 µm, with the peak of spectrum at around 0.5 µm.
Thermal radiation refers to energy emitted at temperatures typical of the earth's surface, about
300°K. The energy lies in the region 4 ≤ λ ≤100 µm, with the peak of the spectrum at about 10
µm. The incoming thermal radiation refers to that emitted by the component gases of the
atmosphere, such as water vapor and carbon dioxide, and other so-called greenhouse gases. The
outgoing thermal radiation is the energy emitted by the ground. Because, the ground is usually
warmer than the atmosphere, the outgoing thermal radiation usually exceeds the incoming
thermal radiation.
The sensible heat flux is the energy flux from the atmosphere to the ground driven by
temperature differences between the ground and the atmosphere. During the daytime, energy
16
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ flows away from the ground into the atmospheric boundary layer, while during the night the
boundary layer supplies energy to the ground.
The latent heat flux refers to the energy used to evaporate moisture from the ground. The soil
heat flux refers to the energy that is supplied to the ground, and which ultimately determines the
temperature of the soil layer.
By definition, the net radiation, RN , is the difference between the solar radiation absorbed at the
surface, and the net thermal radiation emitted by the surface. is the sensible heat flux
supplied to the boundary layer, is the latent heat flux related to the evaporation of water
from the surface, and G is the heat flux into the soil.
SH
LH
During the day, is usually greater than zero: heat is supplied to the atmosphere. During the
night, < 0; heat is drawn from the atmosphere and the ground to support the cooling of the
ground as
SH
SH
RN becomes negative. The cooling can be inhibited in the presence of clouds which
radiate energy towards the ground.
When the ground is moist, most of the incoming radiation can go towards evaporation. An
approximate method of accounting for energy going into evaporation is to assume that the ratio
of latent heat flux to sensible heat flux is a number, referred to as the Bowen ratio, that depends
only on the type of surface being considered. In the next section, we consider the methods used
in AERMET to calculate the components of the surface energy balance.
Solar Radiation
In AERMET, the solar radiation reaching the ground is calculated by reducing the solar flux at
the top of the atmosphere – 1350 W/m2 – through several factors that account for different
physical effects. First, the solar radiation flux normal to the ground surface is determined by the
cosine of the zenith angle, which is the angle between the normal to the surface and the direction
of the incident solar flux. The zenith angle is a function of the latitude of the receptor, the time of
day, and the declination angle. The declination angle is the angle between the normal to the plane
of rotation of the earth about the sun and the axis of rotation of the earth. The declination angle
17
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ varies between +23.5o on July 22 (summer solstice) and -23.5o on December 22 (winter solstice)
as the earth rotates about the sun.
The energy flux in the solar beam at the top of the atmosphere is reduced by absorption and
scattering by gases, particles, and clouds as it travels through the atmosphere to the ground. A
fraction of the scattered radiation reaches the ground as diffuse radiation, which adds to the
direct beam solar radiation. In AERMET, these effects are accounted for through semi-empirical
factors that depend on location and cloud cover. It turns out that the maximum solar radiation
reaching the ground is about 1000 W/m2, and this value is multiplied by the cosine of the zenith
angle (sine of the elevation angle) to give the flux per unit area normal to the ground surface.
This value is further reduced by a factor that is a function of cloud cover.
Outgoing and Incoming Thermal radiation
The earth’s surface emits long-wave radiation (4 ≤ λ ≤100 µm), which can be estimated from
the surface temperature, which is usually tens of degrees higher than the temperature at 10 m.
Because the surface temperature is not measured, AERMET, following Holtslag and Van Ulden
(1983)8, assumes that the outgoing thermal radiation is that associated with the temperature at 10
m plus a value that depends on the difference between the surface temperature and the 10 m
temperature. This additional value is then related empirically to the net radiation. Thus, during
the daytime, when the surface temperature is higher than the 10 m temperature, the correction
adds to the value computed with the 10 m temperature. During the night, when the surface is
cooler than the 10 m temperature, the correction subtracts from the radiation calculated with the
10 m temperature.
The incoming long-wave radiation is a complicated function of the vertical profiles of the
temperature and the concentrations of the gases that absorb and emit in the thermal spectrum.
Clouds are major emitters of thermal radiation; downward thermal radiation can increase
substantially in the presence of clouds. In AERMET, the incoming thermal radiation is
8 Holtslag, A.A.M., and A.P. van Ulden, 1983: A simple scheme for daytime estimates of the surface fluxes from routine weather data, J. Clim. Appl. Meteorol., 22, 517-529.
18
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ calculated using empirical expressions that depend on the sixth power of the 10 m temperature
and cloud cover.
Sensible and Ground Heat Fluxes
During the day, the ground heat flux, G , is usually small relative to the sensible heat flux. In
AERMET, it is taken to be a tenth of the net radiation on the basis of observations. This ensures
that during the day G is into the ground because the net radiation is positive; during the night, G
is directed towards the surface because the net radiation is negative.
The difference between the net radiation and the ground heat flux is partitioned between the
sensible heat flux, H, and the latent heat flux, L. AERMET assumes that the ratio of latent heat
flux to sensible heat flux is a number, referred to as the Bowen Ratio, Bo, that depends only on
the type of surface being considered. This relatively simple approximation can be improved
upon by using other techniques, such as the one proposed by Holstlag and Van Ulden (1983).
However, it should be pointed out that none of these methods is particularly reliable because
evaporation at the surface depends on a process, the transport of moisture in the soil, which is
difficult to parameterize. So the final expression for the sensible heat flux, H, becomes
GRBo
HHH NSLS −=⎟⎠⎞
⎜⎝⎛ +=+
11 , (2)
so that
( GR )Bo
BoH NS −+
=1
. (3)
When the surface is dry so that the sensible heat flux is much greater than the latent heat flux, the
Bo is large, and the heat flux is essentially 0.9RN because G=0.1RN. When the surface is
relatively wet, Bo is small. The heat flux is sensitive to the value of Bo when the ratio is about 1.
It is difficult to specify Bo because it can vary substantially at a single location. However,
variables required in air pollution modeling are not directly proportional to heat flux. So we
might be able to get away with small errors in heat flux. We next show how the daytime
boundary layer height is estimated using the surface heat flux.
19
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ Convective Boundary Layer Height (zic)
The height of the mixed layer, zic, is estimated by assuming that the sensible heat input to the
atmosphere is used to modify the potential temperature in the mixed layer. We illustrate the
procedure by considering a mixed layer that grows by eroding a layer with a stable potential
temperature gradient. Note the symbol H, rather than Hs, is used here to denote sensible heat
flux.
Height of the Convective Boundary Layer
Sensible Heat Flux
A B
C
Stable Potential Temperature Gradient
Zi
( )
TCHz
TH21zC
21
dttHzC21
2/1
p
maxic
2
max2icp
T
0icp
⎟⎟⎠
⎞⎜⎜⎝
⎛=
=
∫=
γτρ
τγρ
θΔρ
Assume that the initial potential temperature at sunrise is stable, and can be represented by the
profile AC shown in the figure. The upward surface heat flux after sunrise results in the
modification of the temperature of the boundary layer. Over a time period, T, the temperature in
the boundary layer changes from AC to BC. Note that the potential temperature (not the actual
temperature) is uniform over most of the boundary layer. The upward heat flux at the surface is
driven by a temperature difference between the surface and the mixed layer, but this temperature
gradient is confined to a shallow layer relative to the mixed layer height.
The energy change in the boundary layer caused by the sensible heating at the surface
corresponds to the area of the triangle ABC. Denoting the potential temperature gradient of AC
by γ , and the temperature change AB by Δθ , this energy change can be written as
icp zC21ABC in Energy θΔρ= . (4)
20
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Noticing that iczγθΔ = , we can equate this energy to the sensible heat flux integrated over T to
obtain
( )dttHzC21 T
0
2icp ∫=γρ , (5)
where H(t) is the time-varying sensible heat flux. For simplicity, we assume that the heat flux
varies linearly with time, so that
( )τtHtH max= , (6)
where τ is a convenient time scale, whose value is not important at this point. Substituting
Equation (6) into Equation (5) and integrating we obtain
τ
γρ2THzC
21 2
max2icp = , (7)
which leads to
TCHz
2/1
p
maxic ⎟
⎟⎠
⎞⎜⎜⎝
⎛=
γτρ. (8)
In AERMET, the initial temperature profile at sunrise corresponds to the actual early morning
sounding, so that the potential temperature gradient, γ, above the mixed layer varies with time as
the mixed layer grows. In addition, the surface heat flux, computed from the surface energy
balance, generally increases from sunrise to noon and then decreases to zero at sunset.
Although the simple model considered here does not incorporate these time variations, it
provides insight into the variables that control the mixed layer height in AERMET. Notice that
the mixed layer height, zic, is proportional to the square root of the heat flux; so errors in
estimating the heat flux are dampened in this calculation. We have a similar situation with the
potential temperature gradient: . So the calculation of z2/1ic /1z γ∝ ic is forgiving of errors in
specifying the potential temperature profile in the morning. The simple formula also tells us that
zic can become very large if the potential temperature gradient is close to zero.
21
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ The background of the last two sections allows an examination of the convective velocity scale,
w*.
Convective Velocity Scale (w*)
The convective velocity scale, w*, is a measure of the turbulent velocities created by surface
heating, referred to as convection. To see this, we will examine a simple model for the motion of
an air parcel that becomes buoyant after heating at the surface. Assume that the mass of the
parcel is 1 kg, and it acquires a temperature excess of T ′ over its surroundings at temperature T.
This results in an upward buoyant force of T/Tg ′ , which accelerates the air parcel upwards. If
this force acts over a distance z, we can estimate the velocity, w, acquired by the air parcel by
equating the work done by the force to the kinetic energy of the parcel at z,
2w~zTTg ⎟⎠⎞
⎜⎝⎛ ′
. (9)
The left hand side of the equation is the product of the buoyant force and the distance, z, over
which it acts, and the right hand side of the equation is the kinetic energy of the parcel with unit
mass.
We can introduce the sensible heat flux into the equation by multiplying both sides of the
equation by w to obtain
3w~zTTwg ⎟⎠⎞
⎜⎝⎛ ′
. (10)
Now the term represents the temperature excess being carried upwards by the air parcel. In
fact, if multiply the term by ρC
Tw ′
p we find that
p
S
Sp
CH
Tw
or
HTwC
ρ
ρ
~
~
′
′
. (11)
The combination HS/ρCp is called the kinematic heat flux, and has units of velocity multiplied by
temperature, and is used instead of the sensible heat flux in constructing micrometeorological
22
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ variables. If we substitute Equation (11) into Equation (10), we can define the free convection
velocity, uf
3/1
0⎟⎟⎠
⎞⎜⎜⎝
⎛≡ z
CH
Tgu
p
Sf ρ
, (12)
where T0 is a reference temperature, taken to be the near surface value, and HS is the surface
sensible heat flux. Notice that uf depends on the 1/3 power of the heat flux, which suggests that
errors in estimating heat flux are reduced in estimating uf.
Observations indicate that the turbulent velocity, σwc, associated with buoyancy production is
given by
icfwc 0.1zz for u3.1 ≤=σ . (13)
Above z>0.1zic, σwc is relatively constant,
, (14) ∗=
==
0.6w
0.1zz at evaluated u3.1 icfwcσ
where the convective velocity scale, w*, is defined by
3/1
0⎟⎟⎠
⎞⎜⎜⎝
⎛≡∗ ic
p
S zC
HTgwρ
. (15)
So the convective velocity scale, w*, is an estimate of the turbulent velocity created by buoyancy
or free convection. Let us estimate its value for a mixed layer height of 1000 m, and a surface
heat flux of 200 W/m2. Taking ρCp=1200 J/(m3K), and T0=300 K, we find w*=1.76 m/s. So in
the upper part of the convective boundary layer, the turbulent velocities are of the order of 1 m/s.
Even at 10 m, the velocity estimated from Equation (15) is about 0.5 m/s.
It turns out that the standard deviation of the horizontal turbulent velocity fluctuations, σv, is
about 0.6w* through the depth of the boundary layer.
23
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ Surface Friction Velocity (u*) and Mechanical Mixed Layer Height (zim)
Air flowing over a surface exerts a shear stress that depends on the level of turbulence in the
boundary layer. We can establish the relationship between the turbulent velocities, σw and σv, and
the shear stress at the surface, τ0, by defining the surface friction velocity as follows:
ρτ0u ≡∗ , (16)
where ρ is the density of air.
When the boundary layer is stable or neutral, turbulent velocities close to the surface are related
to the surface friction velocity through
∗
∗
==
u2u3.1
vs
ws
σσ
, (17)
where we have included ‘s’ in the subscripts to emphasize the fact that these turbulent velocities
are generated by wind shear.
As indicated earlier, turbulence in the stable boundary layer is generated by wind shear, and
inhibited by the stable potential temperature gradient. Observations indicate that the height of
the boundary layer, which is the height to which the turbulence extends, is related to the surface
friction velocity. AERMET uses the relationship for the mechanically generated boundary layer
, (18) 2/3im Auz ∗=
where A is an empirical constant. We see that the boundary layer height is sensitive to the surface friction velocity.
The vertical turbulent velocity in the stable boundary layer decreases with height as follows:
( )2/1
imws z
z1u3.1z ⎟⎟⎠
⎞⎜⎜⎝
⎛−= ∗σ . (19)
The mean velocity during near neutral conditions is also related to the surface friction velocity
through the well-known logarithmic profile:
24
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛ −= ∗
0zhzln
kuzU , (20)
where k is the von-Karman constant=0.4, z0 is the roughness length, and h is the displacement
height. The roughness length, z0, is a function of the physical height of the obstacles on the
surface over which the air flows, and d=5z0. AERMET modifies Equation (20), described later,
to account for the effects of surface heat flux on the surface friction velocity, u*.
Monin-Obukhov Length (L)
The free convection velocity, uf, is a measure of the turbulent velocities generated by buoyancy.
Equation (12) tells us that uf increases with z. On the other hand, the turbulent velocities
generated through shear are relatively constant in the lower tenth of the boundary layer. The
height at which turbulence levels generated by buoyancy are comparable to those generated by
shear is the Monin-Obukhov length, L, obtained by equating uf evaluated at z = L to u*,
30
*
3/1
0
~
~
∗
⎟⎟⎠
⎞⎜⎜⎝
⎛
uHC
gT
L
or
uLC
HTg
S
p
p
S
ρ
ρ
. (21)
We have not used an equal sign in the expression because, for the moment, we want to
emphasize the physical meaning of the Monin-Obukhov length. We see that shear production
dominates buoyancy production of turbulence below the height L. Buoyancy production
dominates above L. So when z<<L, u* governs turbulent velocities, and the boundary layer is
essentially neutral at these heights. The boundary layer is convective for z>>L.
The M-O length, L, is formally defined as
30∗−≡ u
kHC
gT
LS
pρ, (22)
where k is the von-Karman constant. Because HS > 0 during the daytime, L is negative during the
day. During the night, the heat flux is negative, so that L is positive. The physical interpretation
discussed earlier holds for the absolute value of L.
25
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ Before we discuss how L is used in the AERMOD interface, we show how we combine the
turbulent velocities corresponding to buoyancy and shear.
Estimating turbulent velocities in the boundary layer
We computed the turbulent velocities corresponding to the situations when buoyancy or shear
dominates turbulence production. To account for the fact that both mechanisms act together in
the real boundary layer, AERMET combines the velocities using a method that we illustrate for
computing σw
, (23) ( 3/13ws
3wcw σσσ += )
If we substitute Equations (14) and (17) for σwc and σws into Equation (23), and using the
definition of L, we get
3/1
w kLz1u3.1 ⎟⎠⎞
⎜⎝⎛ −= ∗σ . (24)
We see that σw works out to be the neutral value multiplied by a function of z/L. Similar
expressions are used for the turbulent velocities in the horizontal directions. This technique of
correcting for the effects of stability using functions of z/L forms the basis of the similarity
methods used to construct the mean profiles.
Monin-Obukhov (M-O) Similarity
The AERMOD interface uses the micrometeorological variables estimated using AERMOD to
construct profiles of the mean horizontal velocity and temperature, and three components of the
turbulent velocities. These profiles are constructed using M-O similarity, which we have already
encountered in Equation (26). We will illustrate the underlying principles of the theory by
considering the mean velocity profile in the stable boundary layer.
Consider the logarithmic velocity profile in the neutral boundary layer:
( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛ −= ∗
0zhzln
kuzU , (25)
which can be rewritten as
26
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
( )⎟⎟⎠
⎞⎜⎜⎝
⎛ −=
∗ 0zhzln
k1
uzU . (26)
Equation (26) tells us that velocity profiles in all neutral boundary layers, governed by different
roughness lengths, are similar if the wind speed is normalized by the velocity scale, u*, and the
height (z-h) by the length scale z0. This is one statement of similarity theory; the literature
provides other interpretations all of which reduce to the hypothesis that a small number of
velocity, length, time, and temperature scales can be used to summarize the mean and turbulent
structure of the atmospheric boundary layer.
When buoyancy effects are important, the additional length scale is the M-O length, L. For
example, under stable conditions, the velocity profile can be described by
( )⎥⎦
⎤⎢⎣
⎡ −+⎟⎟
⎠
⎞⎜⎜⎝
⎛ −=
∗ Lzz
zhzln
k1
uzU 0
0β . (27)
The velocity profile in the unstable boundary layer can be described using a different function of
z/L.
The temperature profile can be described in a similar manner using a temperature scale, T*,
defined by
∗
−≡uC
HT
p
S
ρ* , (28)
which can be used to describe the temperature profile in the stable boundary layer,
( )⎥⎦
⎤⎢⎣
⎡ −+⎟⎟
⎠
⎞⎜⎜⎝
⎛ −=
−
∗ Lzz
zhzln
k1
TTzT 0
0
0 β , (29)
where T0 is an effective surface temperature.
Equations such as (27) and (29) apply to heights of the order of the M-O length, L, which defines
the extent of the surface boundary layer. Above the surface layer, the relevant scale is the
boundary layer height, zic or zim, and profiles of mean and turbulent variables become functions
of z divided by the boundary height.
27
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ Analysis of data obtained from field studies conducted over the last 30 years has resulted in
semi-empirical formulations to describe the structure of the boundary layer. The AERMOD
interface uses these formulations to construct the profiles required by AERMOD. The inputs are
the parameters: z0, L, u*, w*, zim, zic, U(z), and T(z) contained in the surface file produced by
AERMET. If actual measurements of velocity, temperature, and turbulence are available at
different heights in the boundary layer, the AERMOD interface can incorporate them in
constructing these profiles.
28
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
4 Methodology
The general method used to create the AERMOD *.sfc and *.pfl input files is described in this
section. Appendix B provides “readme files” for each sub-area that contain further details
pertaining to the data and processing procedure used for each sub-area individually.
Data
The data used to produce the AERMOD *.sfc and *.pfl input files are the hourly near-surface
values measured at sites across Southern California for the following meteorological variables:
Wind speed and direction (near-surface)
Temperature (near-surface)
Solar radiation (Surface, Watts per Meter-Squared)
Fractional cloud coverage
These data, along with the temperature profiles from the standard 12-hourly NWS rawinsondes,
are used to produce the *.sfc and *.pfl files.
Table 4.1 summarizes the time period used to construct the input files for each sub-area. The
time period for most sub-areas is 2005 – 2007. For some sub-areas, as noted in the table, the data
are only available or of sufficient quality to allow processing for a shorter time period.
Table 4.2 lists the AQMD monitoring station coordinates, elevations and heights of their wind
and temperature measurements. The measurement heights are based on descriptions of the
towers provided to us by AQMD.
Meteorological Stations
The data for the above variables were obtained from AQMD, National Weather Service (NWS)
and California Irrigation Management Information System (CIMIS) surface meteorological
stations in the AQMD area. Table 4.3 summarizes which data sources were used for which
variable, listed in sequence according to general order of preference. A map of the available
29
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ AQMD, NWS and CIMIS stations in the AQMD area is shown in Figure 4.1. As seen, most
district sub-areas contain one or more of these meteorological stations.
The Coachella Valley was divided into two sub-areas by the arbitrarily drawn blue line shown in
Figure 4.1. One sub-area corresponds to the Palm Springs AQMD station (PLSP) and the other
to the Indio AQMD station (INDI). The San Bernadino (SNBO) was also divided into two sub-
areas, one corresponding to the Fontana AQMD station (FONT) and the other to the San
Bernadino AQMD station (SNBO). This division is denoted in Figure 4.1 by the blue line drawn
arbitrarily across the AQMD-defined SNBO sub-area.
We did not create *.sfc and *.pfl files for sub-areas that did not have an AQMD station. When
wind data was missing at a particular AQMD station, we did not fill the data from stations in
adjacent sub-areas because wind speeds and directions can vary considerably from sub-area to
sub-area. On the other hand, when temperature, solar radiation or cloud cover were missing or
unusable, data from stations in adjacent or nearby sub-areas were used because these variables
vary much less in space. As a result of this procedure, the only missing data in the hourly
meteorological data input to AERMET were missing winds.
All data used in creating the AERMET input files were inspected to make sure that they were
within reasonable bounds using procedures described in the Phase I Interim Report.
In the discussion that follows, each sub-area will be referred by the name of the AQMD station
used as the data source. For example, the sub-area containing the LAXH AQMD station will be
called the ‘LAXH sub-area’.
Data Processing
The following describes the general procedure applied for all sub-areas to process the raw
meteorological data into “pre-processed” files to be read into AERMET. Further details of the
data and processing procedure for each sub-area are given in the ‘readme’ files in Appendix B.
30
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ (i) Wind Speed and Direction
Most sub-areas contain one AQMD surface monitoring station, from where the wind data for
each sub-area were taken. The ‘AZUS’ sub-area contains two stations, AZUS and GLEN, from
which we chose AZUS to characterize the sub-area. No *.sfc and *.pfl input files were produced
for sub-areas that did not contain an AQMD monitoring station.
The wind data at each AQMD station were supplied to us by AQMD. These data were then
inspected by us to identify isolated hours or periods during which winds were obviously
inaccurate or questionable. Table 4.4 lists these time periods and describes the steps taken by us
to treat them.
The wind data were then inspected by us to separate instances of reported zero wind speed and
zero wind direction (‘zero, zero’) in the AQMD wind data files into two groups: those that we
judged “calm” and those that we judged “missing”. This was necessary since the AQMD data
indicated ‘zero, zero’ for both hours of missing wind data as well as during hours when the wind
was evidently not fast enough to register any readings (“calm”). Our judgment of what was
“missing” and what was “calm” was subjective, but followed some basic guidelines:
Isolated instances (periods of less than around 4 hours) of ‘zero, zero’ were considered
“calm” when the non-zero wind speeds reported for hours on either side of the ‘zero,
zero’ period were weak (around 1 m/s or less). This often occurred during nighttime
hours.
Isolated instances (periods of less than around 4 hours) of ‘zero, zero’ were considered
“missing” when the non-zero wind speeds reported for hours on either side of the ‘zero,
zero’ period were not weak. These values were replaced with wind speeds and wind
directions obtained through time interpolation of wind data two hours before and two
hours after the missing period. For single hours of such ‘zero, zero’ instances, the time
interpolation procedure followed the EPA “objective procedure” described in Atkinson
31
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
and Lee (1992)9. For more than a single hour of such ‘zero, zero’ instances, the time
interpolation followed considerations discussed in the EPA “subjective procedure”
described in Atkinson and Lee (1992). Our subjective interpolation procedure still used
the two hours before and two hours after the missing period, however the interpolation
weights assigned to each of these four values were subjectively set by us so that the final
interpolated values maintained reasonable wind direction time trends and variability10.
Long periods of ‘zero, zero’ were flagged as “missing”. While there may in fact be hours
within such ‘zero, zero’ blocks that would be better characterized as “calm”, we could not
identify them because the non-zero measurements on either side of the ‘zero, zero’ block
are too far away in time to make a reliable determination of any “calm” hours within the
‘zero, zero’ block. Furthermore, there is no reliable way to fill long periods of missing
winds since the non-zero winds on either side of the ‘zero, zero’ block are too far away in
time to allow for reliable time interpolation.
The characterization of the ‘zero, zero’ periods is presented in Excel spreadsheets, which are
provided for each AQMD station on the accompanying CD under the “WINDGAPS” directory
(Appendix D). Green shaded cells indicate ‘zero, zero’ entries judged “missing”, with the values
in these cells replaced with the missing value indicator ‘-999.0’ for wind speed and wind
direction. Yellow shaded cells indicate ‘zero, zero’ entries judged “calm”, with the values in
these cells left as ‘zero, zero’.
There were also entries in the AQMD wind files that had zero wind speed and non-zero wind
direction. Upon inspection, these entries appeared to correspond to periods of weak wind speeds.
We therefore judged these periods as “calm” and left them untreated because AERMET will
interpret a zero wind speed (regardless of wind direction) as “calm” and not process the hour.
9 Atkinson, D. and R. F. Lee, 1992: “Procedures for Substituting Values for Missing NWS Meteorological Data for Use in Regulatory Air Quality Models”, July 7, 1992; Available at http://www.epa.gov/scram001/metguidance.htm. 10 Scalar wind direction averaging was used following the procedure outlined in the EPA document “Meteorological Monitoring Guidance for Regulatory Modeling Applications”, February 2000, EPA-454/R-99-005; available at http://www.epa.gov/scram001/metguidance.htm.
32
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ The wind roses for the final wind data applied for each sub-area are shown in Appendix A, and
are provided in individual electronic files on the accompanying CD.
(ii) Temperature
The primary source for near-surface air temperature for each sub-area was the AQMD data. If
these data were not available, the temperature data were taken from a CIMIS or NWS station
within the sub-area, or from an AQMD, CIMIS or NWS station from an adjacent or nearby sub-
area.
Short periods of missing temperature data (less than around 4 hours) were filled by time-
interpolation. The interpolation technique used the average of the two non-missing temperatures
immediately before and after the missing period to replace all missing temperatures within the
period. This method is reasonable for the cases that occurred most frequently, when only one or
two hours were missing.
For long periods of missing data, temperatures were taken from an alternative AQMD or CIMIS
station within the sub-area, or from an AQMD, CIMIS or NWS station adjacent or nearby sub-
area. If no suitable station was available, time interpolation was applied.
As a result of this filling procedure, all hours of temperature are ultimately filled.
Details on the temperature data used for each sub-area are given in the ‘readme’ files. These are
printed out in Appendix B.
(iii) Solar Radiation
For most sub-areas, the primary source for surface solar radiation for each sub-area was the
CIMIS data within the sub-area. If these data were not available, the primary solar radiation data
were taken from another CIMIS station within the sub-area, or from a CIMIS station from an
adjacent or nearby sub-area.
Short periods of missing solar radiation data (less than around 4 hours) were filled by time-
interpolation. The interpolation technique used the average of the two non-missing solar
33
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ radiation values immediately before and after the missing period to replace all missing solar
radiation values within the period. This method is reasonable for the cases that occurred most
frequently, when only one or two hours were missing.
For long periods of missing data, solar radiation was taken from an alternative CIMIS station
within the sub-area, or from a CIMIS station from an adjacent or nearby sub-area. In the case of
the ‘LGBH’ sub-area, solar radiation data from the LGBH AQMD station were used to fill long
periods of missing data. Otherwise, if no suitable station to provide solar radiation data was
available, time interpolation was applied.
For the SCLR, RESE, BURK and ELSI sub-areas, we allowed AERMET to calculate solar
radiation based on time of day/year and the input value for fractional cloud coverage. This
appeared to give more realistic results for micrometeorological variables computed by AERMET
compared to what were obtained when using the local CIMIS radiation data.
As a result of this procedure, all hours of solar radiation were ultimately filled.
Details on the solar radiation data used for each sub-area are given in the ‘readme’ files. These
are printed out in Appendix B.
(iv) Cloud Coverage
The primary source for fractional cloud coverage for each sub-area was the NWS data within the
sub-area. Absent these data, the primary cloud coverage data were taken from an alternative
NWS station within the sub-area, or from a NWS station from an adjacent or nearby sub-area
NWS cloud coverage is expressed by codes “CLR” (clear), “FEW” (few), “SCT” (scattered),
“BKN” (broken), and “OVC” (overcast). The translation of these codes to fractional cloud
coverage values follows from recommended values contained in Table 3 of the ASOS User’s
Guide11
11 “Automated Surface Observing System (ASOS) User’s Guide”, NOAA, March 1998; obtainable at http://www.nws.noaa.gov/asos/pdfs/aum-toc.pdf.
34
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Based on this, we assigned the following fractional values (in tenths): “CLR” = 0, “FEW” = 1.0,
“SCT” = 4.0, “BKN” = 8.0, “OVC” = 10.0.
Short periods of missing cloud coverage data (less than around 4 hours) were filled by time-
interpolation. The interpolation technique used the average of the two non-missing cloud
coverage values immediately before and after the missing period to replace all missing
temperatures within the period. This method is reasonable for the cases that occurred most
frequently, when only one or two hours were missing.
For long periods of missing data, cloud coverage values were taken from an alternative AQMD
or CIMIS station within the sub-area, or from an AQMD, CIMIS or NWS station adjacent or
nearby sub-area. If no suitable station was available, time interpolation was applied.
For the CRES sub-area, we applied a default fractional cloud coverage value of 5.0 (tenths) for
all hours since no suitable NWS station was available in this mountainous area.
As a result of this procedure, all hours of fractional cloud coverage are ultimately filled.
Details on the cloud coverage data used for each sub-area are given in the ‘readme’ files. These
are printed out in Appendix B.
35
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ (v) Land Surface Characteristics
Values for three land surface parameters – minimum (“noontime”) surface albedo, Bowen Ratio
and surface roughness length – are inputs to AERMET. Given the uncertainty in determining
these values, we generated two versions of *.sfc and *.pfl files – the first version from surface
parameter values for each sub-area computed using the EPA module AERSURFACE12, and the
second version from surface parameter values that are uniform for all sub-areas. The *.sfc and
*.pfl files for both sets are contained on the accompanying CD in file directories entitled
‘Version 1’ and ‘Version 2’ (Appendix D).
Further details of the choice in surface parameters are now discussed.
Version 1: Land Surface Parameters Determined by AERSURFACE
AERSURFACE determines the values of the three land surface parameters based on the 30-
meter gridded land-use data in the USGS NLCD92 dataset. AERSURFACE first determines the
land-use values around the primary meteorological tower. Look-up tables programmed into
AERSURFACE then map the land-use categories around this tower to values for the three land
surface parameters at each 30-meter grid square. These look-up tables are based on values from
field studies in various locations provided in the literature. The final surface parameter values for
AERMET are then determined by averaging the 30-meter gridded values over a radius of
influence around the tower. The averaged values for surface parameters can be made temporally
and directionally dependent through user-specified settings in AERSURFACE. Further details of
this procedure can be found in the EPA AERSURFACE User’s Manual.
We chose the “average” soil moisture and “arid” surface characteristic settings for running
AERSURFACE, as these seemed most appropriate to arrive at typical (climatologically
averaged) values for Southern California. For simplicity, we specified no time or directional
dependence, and therefore single values for surface parameters output by AERSURFACE were
applied in our AERMET runs for all hours and all wind directions. This simple approach seems
12 http://www.epa.gov/scram001/dispersion_related.htm#aersurface.
36
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ justified due to the large inherent uncertainty in AERSURFACE associated with its
determination of surface parameter values at a user-specified site based on look-up tables of
“typical” values from observations at other sites.
We applied the January 2008 version of AERSURFACE, which was the most current at the time
of this writing. This was done through use of the GUI front-end for AERSURFACE provided in
the Lakes Environmental software AERMET-View13.
The resulting values applied for surface albedo and roughness length for each sub-area output
from AERSURFACE are listed in Table 4.5.
For Bowen ratio, we overrode the values output by AERSURFACE with a specified values for
each sub-area. The values we specified alongside the values output by AERSURFACE are listed
in Table 4.5. For most sub-areas we chose a value of 1.0, except for BNAP, PLSP and INDI
where we chose a value of 1.5 to be closer to the value output by AERSURFACE for these sub-
areas. As seen, AERSURFACE produced in almost all cases a higher Bowen ratio value than the
value we specified. This replacement with a lower value of Bowen ratio is justified for several
reasons. First, the uncertainty in the AERSURFACE look-up table approach is likely to be large
in the case of Bowen Ratio, since latent heat flux is difficult to measure/compute accurately.
Also, the use of NLCD92, rather than the more current 2001 or 2006 sets14, raises doubts about
whether the Bowen ratios output by AERSURFACE for Southern California account for recent
residential developments in many areas of the air basin, where Bowen ratios are likely to be
lower because of dry land with irrigated residential areas. The value of 1.0 set for most sub-areas
therefore seems like a suitable constant “place-holder” value until these issues can be resolved.
Version 2: Uniform Land Surface Parameters Across All Sub-Areas
After consultation with AQMD staff, a second version of the *.sfc and *.pfl files were
constructed using uniform values of surface parameters across all sub-areas. The values used for
each parameter are
13 http://www.lakes-environmental.com/14 http://www.epa.gov/mrlc/nlcd-2006.html
37
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
• Surface Roughness Length = 0.4 m
• Bowen Ratio = 1.3
• Minimum (noontime) surface albedo = 0.2
These values were chosen by AQMD to approximately reflect the median of the values produced
by AERSURFACE for all sub-areas, which are shown in Table 4.5
Application of AERMET
By applying the procedure described above, we produced “pre-processed” meteorological input
files for each sub-area that contain the hourly values of the four meteorological variables over
the time period of interest for each sub-area (see Table 4.1). These files were then read in to
AERMET to produce the *.sfc and *.pfl files needed to run AERMOD for each sub-area.
As discussed in detail in Section 3, AERMET computes the following micrometeorological and
boundary-layer variables needed for characterizing turbulent diffusion in AERMOD:
• Surface friction velocity (u*, meters per second)
• Free convective scaling velocity (w*, meters per second)
• Monin-Obukhov length (L, meters)
• Daytime (“convective”) boundary layer height (zic, meters)
• Nighttime (“mechanical”) boundary layer height (zim, meters)
To compute these, an estimate of the morning vertical temperature gradient within the boundary
layer (γ, Kelvin per meter) is needed, which AERMET calculates based on the standard NWS
morning rawinsonde sounding. We used the Miramar AFB (San Diego area) sounding data to
AERMET for this purpose.
38
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ In addition to calculating the above variables, AERMET also passes through the hourly values of
wind speed, wind direction, temperature for use in AERMOD. Hourly values for the vertical
temperature gradient above the boundary layer (‘Field 9’ in the sample file shown in Figure 2.2)
are also output for use in the plume rise calculations in AERMOD.
To summarize, the main input and output files provided in the deliverables for running
AERMET for each sub-area are the following:
• The “pre-processed” hourly meteorological input file for the sub-area (*.prn).
• The Miramar AFB rawinsonde data (file ‘NKX.ua’)
• AERMET “control” input files (IN1, IN2 and IN3) for the sub-area, which contain
various run parameters as well as the values for surface parameters.
• The *.sfc and *.pfl output files from AERMET, which are, in turn, the meteorological
input files to AERMOD.
We also provided other intermediary and output files produced by AERSURFACE and
AERMET, which contain run-time messages and other similar content.
All of the above files are contained for each sub-area on the deliverable CD (Appendix D).
We applied AERMET version 06341, which is the most current version at the time of this
writing. AERMET was run through use of the GUI front-end provided in the Lakes
Environmental software AERMET-View15.
The meteorological input files produced by AERMET were checked through a code produced by
us to ensure that: 1) micrometeorological outputs were within plausible ranges, and 2) No errors
were made in applying AERMET. This code contains the primary equations in AERMET, and is
designed to accept NWS and AQMD onsite observations directly. The results from the code
15 http://www.lakes-environmental.com/
39
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ were compared with those produced by AERMET. We did not find discrepancies between the
two sets to indicate errors in the application of AERMET.
Post-Processing (Optional)
We have developed three postprocessors that allow modification of the surface files (.sfc)
produced by AERMET to account for processes not included in AERMET, and to examine the
sensitivity of the outputs to surface parameters without rerunning AERMET:
• Adjust_Mxht
• Adjust_SFC_File
• Recompute_SFC_File
• Convert_Rural_to_Urban
The processor called ‘Adjust_Mxht (SFC_File, max_mxht, remark)’ reads the .sfc file, replaces
all the daytime mixed layer heights that exceed a user specified maximum value by this
maximum value, recalculates w*, and writes a new .sfc file. The inputs to the program are the
name of the SFC file (SFC_File), the user-specified maximum mixed layer height (max_mxht),
and a remark (remark). The output file is given the name of the input file plus the remark.
This processor can be used to account for the fact that the daytime boundary layer at a shoreline
location corresponds to the convective internal boundary layer that develops when stable air
from the ocean flows on to warmer land. The height of this internal boundary grows as a function
of distance from the source as the upward heat flux from the ground erodes the overlying stable
layer. Within 10 km from the shoreline, the boundary layer height can be shallow relative to the
height far inland where shoreline effects are absent. In Wilmington, for example, the internal
boundary layer height rarely exceeded 200 m at a distance of 5 km from the shoreline in
40
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ September (Yuan et al. 2006)16. AERMET would instead typically estimate a maximum
boundary layer height of around 1000 m because it does not account for two-dimensional
shoreline is effects. This processor can also be used to limit the mixed layer height to a specified
observed value, which we might be necessary in the South Coast Air Basin where large scale
subsidence can result in lower values of the mixed layer height than those estimated by
AERMET. An overestimated boundary layer height will underestimate concentrations from both
surface as well as elevated releases.
A processor called ‘Adjust_SFC_File (SFC_File, z0_new, Bo_new,remark)’ is designed to
examine the sensitivity of the .sfc information to changes in the Bowen ratio, Bo, and the surface
roughness length, z0. The processor reads the .sfc file, recalculates the heat flux, Qs, the
convective mixed layer height, zic, the surface friction velocity, u*, and the corresponding
mechanical mixed layer height, zim. The formulations used to recalculate these parameters differ
slightly from those used in AERMET; so the adjusted file is not identical to what AERMET
would produce with the new surface parameters. However, the output from this post-processor is
adequate to examine the impact of changes in Bo and z0 on AERMET output as well as on
concentrations. Figure 4.2 compares outputs from the processor with the corresponding
AERMET output for z0=1 m and Bo=2. The processor generated the output from a .sfc file for
z0=0.387 m and Bo=1. We see that there are minor differences in the two outputs especially for
w*. This is because the post-processor neglects changes in the potential temperature gradient
above the mixed layer with height, which results in different mixed layer height and hence w*.
The inputs to the program are:
1. SFC_File: the original file that is adjusted;
2. z0_new: the new roughness length;
3. Bo_new: the new Bowen ratio;
4. remark: a string descriptor of the run.
16 Yuan, J., A. Venkatram, and V. Isakov, 2006: “Dispersion from ground-level sources in a shoreline urban area”, Atmos. Environ., 40, 1361-1372.
41
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ The output file has the name of the input file plus the “remark”.
We have also developed a draft version of a postprocessor that allows examination of the
sensitivity of AERMET output to changes in surface albedo in addition to changes in surface
roughness and the Bowen ratio. The inputs to the processor called Recompute_SFC_File
1. The preprocessed (*.prn) used by AERMET;
2. SFC_File: the original file that is adjusted;
3. z0_new: the new roughness length;
4. Bo_new: the new Bowen ratio;
5. alb_new: the new albedo
6. alb_old: the old albedo
7. remark: a string descriptor of the run.
The output file has the name of the input file plus the “remark”.
As in Adjust_SFC_File, the formulations used to recalculate these parameters differ slightly
from those used in AERMET; so the adjusted file is not identical to what AERMET would
produce with the new surface parameters. Recompute_SFC_File also uses the incoming solar
radiation to estimate the thermal radiation corresponding to the old value of the albedo. Then,
the absorbed solar radiation is recalculated using the new value of the albedo, and the new
absorbed solar radiation is added to the thermal radiation (which is not affected by the change in
albedo) to estimate the net radiation. The new sensible heat flux is then estimated using the new
value of net radiation.
The new sensible heat flux is used to adjust the mixed layer height, but the method does not
account for the new value of the potential temperature gradient above the new mixed layer
height. This results in errors in the calculation of the mixed layer height, and hence errors in the
convective velocity scale, w*. These errors are evident in Figures 4.3 and 4.4, which compare
AERMET outputs with outputs from the processor based on anah.sfc for 2007. The differences
in the surface friction velocity are small, but the deviations in w* do not appear to be negligible.
The usefulness of the processor can be evaluated only be running AERMOD using the two
different input files.
42
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ A processor called ‘Convert_Rural_to_Urban (Input_File, Output_File, zou, dh, Ru, wdf)’
allows one to modify an .sfc file constructed using rural surface measurements to an .sfc file that
accounts approximately for urban effects. The inputs to the program are:
1. Input_File: the name of the original .sfc file;
2. Output_File: the name of the adjusted file;
3. zou: the roughness length in m in the urban area;
4. dh: the displacement height in m of the urban area;
5. Ru: the distance in m between the rural and urban stations;
6. wdf: a two element vector that specifies the wind directions in degrees between which the
correction is applied.
This processor uses the method described in Luhar et al. (2006)17 to account for urban effects.
AERMOD has a method (see Cimorelli et al. 200518) to partially account for urban effects
during the night when the “URBAN” option is selected in AERMOD. This method, however,
only accounts for the upward heat flux that occurs when stable air flows from a rural area over a
warmer urban area during the night. It does not account, on the other hand, for the corresponding
increase in roughness when rural air flows over an urban area. Although the surface files
provided to AQMD implicitly account for urban roughness effects because they are constructed
using data from urban measurement sites, it might still be necessary to account for additional
urban effects by (a) invoking the “URBAN” option when AERMOD is run and/or (b) running
the ‘Convert_Rural_to_Urban’ post-processor to account for changes in surface characteristics
that occur between the meteorological measurement site and the source/receptor location.
17 Luhar, A. K., A. Venkatram, and S. Lee, 2006: “On relationships between urban and rural near-surface meteorology for diffusion applications”, Atmos. Environ., 40, 6541-6553
18 Cimorelli, A. J., S. G. Perry, A. Venkatram, J. C. Weil, R. J. Paine, R. B. Wilson, R. F. Lee, W. D. Peters, and R. W. Brode, (2005): “AERMOD: A dispersion model for industrial source applications. Part I: General model formulation and boundary layer characterization”, J. App. Meteor., 44, 682-693.
43
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table 4.1: Data period considered for each AQMD sub-area
Sub-Area Data PeriodANAH 2005 - 2007AZUS 2005 - 2007BNAP 2005 - 2007BURK 2005 - 2007CELA 2006 - 2007*CRES 2005 - 2007CSTA 2005 - 2007ELSI 2005 - 2007FONT 2005 - 2007INDI 2005 - 2007LAHB 2005 - 2007LAXH 2005 - 2007LGBH 2005 - 2007LYNN 2005 - 2007MSVJ 2005 - 2007PERI 2007 **PICO 10/12/2005 - 2007***POMA 2005 - 2007PLSP 2005 - 2007RDLD 2007 ****RESE 2005 - 2007RIVR 2005 - 2007SCLR 2005 - 2007SNBO 2005 - 2007UPLA 2005 - 2007WSLA 2005 - 2007
* Questionable wind data during 2005 at the CELA station** Questionable wind data during 2005 and 2006 at the RDLD station*** Data only available since 10/12/2005 at the PICO station**** Data only available since 2007 at the PERI station
44
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table 4.2: Coordinates, elevation and measurement heights of AQMD monitoring stations
Station Latitude Longitude Elevation (m)
Height of Wind Measurement
(m)
Height of Temperature Measurement
(m)
ANAH 33o 49' 50" 117o 56' 19" 41 9.1 5.5AZUS 34o 8' 11" 117o 55' 26" 182 9.1 5.5BNAP 33o 55' 15" 116o 51' 30" 660 9.1 5.5BURK 34o 10' 33" 118o 19' 1" 175 12.2 8.5CELA 34o 3' 59" 118o 13' 36" 87 21.3 17.7CRES 34o 14' 29" 117o 16' 32" 1387 9.1 5.5CSTA 33o 40' 26" 117o 55' 33" 20 9.1 5.5ELSI 33o 40' 35" 117o 19' 51" 406 9.1 5.5FONT 34o 6' 1" 117o 29' 31" 367 9.1 5.5INDI 33o 42' 30" 116o 12' 57" -4 9.1 5.5LAHB 33o 55' 31" 117o 57' 8" 82 9.1 5.5LAXH 33o 57' 15" 118o 25' 49" 42 9.1 5.5LGBH 33o 49' 25" 118o 11' 19" 30 12.2 8.5LYNN 33o 55' 44" 118o 12' 39" 29 9.1 5.5MSVJ 33o 37' 49" 117o 40' 30" 170 9.1 5.5PERI 33o 47' 20" 117o 13' 40" 442 9.1 5.5PICO 34o 00' 37" 118o 4' 7" 58 9.1 5.5POMA 34o 4' 0" 117o 45' 0" 270 9.1 5.5PLSP 33o 51' 10" 116o 32' 28" 171 9.1 5.5RDLD 34o 3' 32" 117o 8' 52" 481 9.1 5.5RESE 34o 11' 57" 118o 31' 58" 228 12.2 8.5RIVR 34o 0' 2" 117o 24' 55" 250 9.1 5.5SCLR 34o 23' 0" 118o 31' 42" 375 9.1 5.5SNBO 34o 6' 24" 117o 16' 25" 305 9.1 5.5UPLA 34o 6' 14" 117o 37' 45" 379 9.1 5.5WSLA 34o 3' 2" 118o 27' 24" 97 9.1 5.5
45
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table 4.3: Data sources used for each meteorological variable
Variable Data Source
Wind speed and direction AQMD
Temperature AQMD, CIMIS, NWS
Solar Radiation CIMIS, AQMD, Calculated based on cloud
coverage and time of day
Fractional Cloud Coverage NWS
46
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table 4.4: Instances of potential problems in AQMD wind data.
Sub-Area Problem Course of Action
BURK Hours of unreasonably high wind speeds on
12/18-12/19/07
Replaced with missing value
indicator
CELA Frequent periods of unreasonably high wind
speed from 1/1-10/4/05
Sub-area only processed for 2006 –
2007.
CRES A few hours of slightly negative wind speed
Effectively set to zero since we
entered to wind speeds to one
decimal place into AERMET.
CSTA Frequent hours of very light winds Not treated
ELSI Unreasonable wind speeds on 1/10-1/14/05 Replaced entire period with missing
value indicator
FONT Unreasonable wind data during period 5/3-
8/2/06
Replaced entire period with missing
value indicator
INDI One hour of slightly negative wind speed
(1/24/05)
Effectively set to zero since we
entered to wind speeds to one
decimal place into AERMET.
LAHB Frequent hours of very light winds Not treated
LAXH
Hours of unreasonably high wind speeds
(9/26/05, 4/20/06, 7/29/06, 11/18/06,
11/9/07, 11/11-11/12/07)
Replaced with missing value
indicator
47
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
LYNN Frequent hours of very light winds Not treated.
MSVJ
Frequent hours of questionably high wind
speeds often occur around sunrise and
sunset hours
Not treated
POMA
Frequent hours of questionably low wind
speeds from 1/29 – 7/10/07 (long periods of
0.3 and 0.2 mph).
Not treated
RDLD
Lots of zero and negative wind speed values
during 2005 to August 2006.
Frequent hours of very light winds.
Sub-area only processed for 2007.
Not treated.
RESE
Frequent hours of very light winds
Hours of unreasonably high wind speeds
(12/26/05, 5/22/06, 10/13/07).
Not treated
Replaced with missing value
indicator
SCLR Questionably low wind speeds on 3/29 –
4/4/06 Not treated
UPLA Questionably low wind speeds on 7/20 –
8/1/06 Not treated
48
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table 4.5: Surface parameter values applied to AERMET.
Sub-Area Surface Albedo
Surface Roughness
(meters)
Bowen Ratio
Bowen Ratio from AERSURFACE
ANAH 0.17 0.524 1.0 1.18AZUS 0.19 0.421 1.0 1.68BNAP 0.22 0.157 1.5 2.32BURK 0.19 0.626 1.0 1.58CELA 0.18 0.645 1.0 1.42CRES 0.17 0.412 1.0 1.35CSTA 0.18 0.403 1.0 1.20ELSI 0.20 0.265 1.0 1.49FONT 0.19 0.273 1.0 1.30INDI 0.19 0.235 1.5 0.98LAHB 0.18 0.554 1.0 1.17LAXH 0.16 0.255 1.0 0.62LGBH 0.18 0.596 1.0 1.35LYNN 0.18 0.505 1.0 1.25MSVJ 0.18 0.351 1.0 1.32PERI 0.20 0.215 1.0 1.24PICO 0.18 0.396 1.0 1.28POMA 0.18 0.548 1.0 1.23PLSP 0.22 0.509 1.5 3.02RDLD 0.20 0.408 1.0 1.53RESE 0.18 0.611 1.0 1.17RIVR 0.19 0.387 1.0 1.34SCLR 0.21 0.291 1.0 2.16SNBO 0.18 0.361 1.0 1.20UPLA 0.18 0.399 1.0 1.15WSLA 0.18 0.451 1.0 1.37
49
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 4.1: Map of South Coast Air Quality District area showing locations meteorological measurement stations
across the AQMD area. Shown are the following: AQMD surface stations, AQMD profiler stations, NWS surface
stations, and CIMIS surface stations.
50
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 4.2: Output from AERMET for z0= 1 m and Bo=2 for Riverside for 2007 compared with output from
Adjust_SFC_File generated from the ‘rivr.sfc’ file output from AERMET for z0= 0.387 m and Bo=1. Top
panel compares variables paired in time while bottom panels compare distributions of outputs.
51
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 4.3: Output from AERMET for z0= 0.1 m and Bo=2, and albedo=0.32 for Anaheim for 2007 compared with
output from Recompute_SFC_File generated from the ‘anah.sfc’ file output from AERMET for z0= 0.524 m, Bo=1,
and albedo=0.19. Top panel compares variables paired in time while bottom panels compare distributions of
outputs.
52
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 4.4: Output from AERMET for z0= 0.4 m and Bo=0.5, and albedo=0.0.16 for Anaheim for 2007
compared with output from Recompute_SFC_File generated from the ‘anah.sfc’ file output from AERMET for
z0= 0.524 m, Bo=1, and albedo=0.19. Top panel compares variables paired in time while bottom panels
compare distributions of outputs.
53
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
5 Results
We present our analysis of the *.sfc and *.pfl files in two steps. First, a basic assessment to check
for typical features of average Southern California meteorology is presented. Second, we
compare results from the Riverside (RIVR) and Long Beach (LGBH) sub-area *.sfc files to
observations. Results are only shown for the ‘Version 1’ files.
Basic Assessment
(i) Wind Speed and Direction
The wind roses from each AQMD monitoring station used to construct the *.sfc files are shown
in Appendix A. A map in which these wind roses are overlaid is shown in Figure 5.1. This map
shows that the basic onshore and topographic channeling flows typical of the prevailing wind
flow pattern across the AQMD air basin appear to be successfully incorporated into the input
files.
An important difference between AERMOD and ISCST3 is that AERMOD can be run for
conditions in which the wind speed is less than 1 m/s. The frequency of wind speeds less than 1
m/s among the AQMD monitoring station data used to create the input files is therefore shown in
Figure 5.2. Relatively bigger differences in concentrations between AERMOD and ISCST3 may
be anticipated in the sub-areas where the percentage of wind speeds less than 1 m/s are relatively
high. As discussed in Section 4 (see Table 4.4), the percentage of low wind speeds at some of
these, however, may be questionable. The lower bound of wind speed in creating the *.sfc and
*.pfl files was 0.1 m/s.
(ii) Micrometeorological Variables
The average surface friction velocity (u*) over hours 0900 – 1500 LST for all sub-areas is shown
in Figure 5.3. These data depict typical daytime values of u*. By inspecting the pattern of values
across the map, it can be seen that many of the coastal and moderately inland sub-areas contain
relatively higher u* than sub-areas that are further inland and in interior valleys. This is expected
54
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ behavior due to the prevalence of sea-breeze flows closer to the coast. Of course, the highest
daytime values of u* occur at BNAP and SCLR, which are prone to very high wind speeds.
The average free convective velocity scale (w*) over hours 0900 – 1500 LST for all sub-areas is
shown in Figure 5.4. These data depict typical values of w*. By inspecting the pattern of values
across the map, it can be seen that many of the coastal and moderately inland sub-areas contain
relatively lower w* than sub-areas that are further inland or in interior valleys. This is expected
due to the stronger heat fluxes further inland due to relatively higher solar radiation and lower
cloud coverage. In the case of BNAP and PLSP, the higher w* is also a result of the relatively
higher specified value of Bowen ratio for these sub-areas. Values, however, are probably too
uniform across the air basin since daytime boundary layer heights appear to be overestimated in
coastal sub-areas, as discussed below.
The average surface friction velocity (u*) over hours 2100 – 0500 LST for all sub-areas is shown
in Figure 5.5. These data depict typical nighttime values of u*. By inspecting the pattern of
values across the map, three “clusters” of values are evident. The cluster of stations (BNAP,
SCLR, PLSP, INDI) with relatively high nighttime u* appear to exhibit the high wind speeds in
the area. Many of the stations in the second cluster of stations (CELA, LGBH, LAXH, ANAH,
PICO, BURK) with average nighttime values around 0.1 m/s are probably exhibiting the
moderate nighttime wind speed conditions in the coastal and near-inland basin due to their
relatively closer coastal proximity. Several of the stations in the third cluster (RDLD, LAHB,
WSLA, POMA, CSTA, LYNN, PERI, RESE, ELSI, SNBO) with relatively low average
nighttime values around 0.05 m/s, on the other hand, are probably exhibiting more localized
effects. This could be a characteristic of their more inland location (PERI, ELSI) or other local
effects that are unknown (WSLA). The relatively low nighttime u* at some of these stations, for
example CSTA, RDLD, POMO, LAHB, could also due to questionable wind data, as discussed
in Section 4 (see Table 4.4).
55
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ (iii) Daytime Boundary-layer heights
Average values of daytime boundary layer height over the hours of 1100 – 1500 LST for each
sub-area are listed in Table 5.1. While inland average values appear reasonable, the uniformity of
values across the basin is questionable. In particular, the coastal sub-areas (LGBH and LAXH)
should not have similar values as those far inland (RIVR, PLSP), but should have noticeably
smaller average daytime boundary layer depths due to the presence of stable internal boundary
layers from sea-breeze intrusion. AERMET does not include a mechanism to include such effects
in its formulation for computing boundary layer height, so the uniformity is therefore expected.
Corrections to include internal boundary layer effects for coastal locations must therefore be
done in post-processing.
Comparison with Observations
Because urban areas are not horizontally homogeneous, it is necessary to evaluate the
applicability of AERMET to urban areas. Here, we address this issue by comparing selected
variables from the AERMET Riverside sub-area rivr.sfc and Long Beach sub-area lgbh.sfc files
with measurements made in two urban areas in the South Coast air basin: Riverside and
Wilmington.
The roughness length, Bowen ratio and minimum surface albedo for Riverside were estimated
with AERSURFACE. The values of these parameters are based on the characterization of the
surface surrounding the AQMD sites at which surface measurements for AERMET were made.
The relationships between surface descriptors and surface parameters used in AERMET are
uncertain. Thus, the values of the surface parameters are uncertain. The Bowen ratio, which is
the ratio of the sensible to the latent heat flux, is likely to be the least reliable of the parameters
because it depends on soil moisture, which depends on soil and rainfall history that cannot be
captured in a static descriptor of the surface. In order to acknowledge this uncertainty, we have
used a nominal Bowen ratio of unity in constructing AERMET files for all the AQMD sites
except at two desert sub-areas (BNAP and PLSP) where latent heat flux is likely to be small. We
have evaluated the use of this nominal value of Bowen ratio by comparing AERMET outputs
with observations made in Riverside and Wilmington.
56
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ The meteorological observations used in the comparison were obtained from Professor Marko
Princevac, UC Riverside. His group made measurements at three sites in Riverside County,
California, in 2007. Site US (upwind suburban) is in a desert plain in Moreno Valley, located
east of Riverside. Site DS (downwind suburban) is on top of a bluff located above the Santa Ana
River in suburban Riverside, located west of Riverside. Site CU (center urban) is located on the
street corner of Arlington and Brockton in downtown Riverside. Sites US and CU are 18 km
apart and sites CU and DS are 9 km apart.
Each site was equipped with a 3 meter tower instrumented with (1) a sonic anemometer, (2) two
soil heat flux plates, (3) an infrared thermometer, (4) a krypton hygrometer, (5) two soil
temperature probes, (6) a water content reflectometer, (7) two air temperature sensors, and (8)
site US had a net radiometer.
Data were collected from early February through late April 2007 at Site CU. Sites US and DS
were run for shorter periods of time during mid-March through late April 2007 and late March to
the end of April 2007, respectively. The comparisons that follow are based on 1-hour averaged
data from the sonic anemometers.
Figure 5.6 compares measurements of the surface variables, sensible heat flux, and the surface
friction velocity measured at the US site with AERMET estimates for the same time. These
variables govern dispersion of a near surface release. The top panels, which compare the values
paired in time, show that heat fluxes and friction velocities estimated at the nominal Riverside
site differ from the measurements at the US site. This is to be expected because the surface
characteristics at the Riverside site differ from those at the US site. In this case, most of the
AERMET friction velocities and heat fluxes are within a factor of two of the observed values at
this site.
The bottom panels compare the variables after ranking them from lowest to highest value; the
values are no longer paired in time. This method compares the distributions of the surface
friction velocities and heat fluxes at the two sites. This comparison is relevant because these
distributions govern the distributions of concentrations, which are important from a regulatory
viewpoint. It answers the question: Do the paired-in-time differences in meteorological variables
translate into differences in the distributions of concentrations. In this particular case, the
57
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ differences are small except at the tails of the distributions: the observed heat fluxes and surface
friction velocities have a few high values that are not present in the AERMET distributions. This
might not have an impact on design concentrations, which are governed by low values of these
variables. Note that AERMET distribution of u* is close to the observed distribution at low
values, suggesting that design concentrations for surface releases are likely to be realistic.
Figure 5.7 compares the diurnal variation of the observed and AERMET variables for 4/26/2007,
a representative day. AERMET follows the diurnal trend of the observed values, although the
peak heat flux occurs about two hours earlier than the observed value. This is a feature that is
observed at all three sites.
The magnitudes of the AERMET surface friction velocities are similar to observed values for
this particular site. The results are similar for 3/31/07 as seen in Figure 5.8.
Figure 5.9 shows the comparison of AERMET estimates with measurements made in the
Riverside downtown area at site CU. Here we see that AERMET overestimates the surface heat
flux by almost a factor of two, but surface friction velocities from AERMET compare well with
observed values. The overestimation of heat flux could be related to the fact that the CU site was
located on a well watered lawn where the Bowen ratio might be smaller than the nominal
AERMET value of 1.
At site DS (Downwind suburban), AERMET underestimates both the surface heat and the
surface friction velocity by almost a factor of two, as seen in Figure 5.10. Note that there are
several observed values of large positive heat fluxes when AERMET values are negative. This
suggests the need for including urban effects when running AERMOD; the AERMOD interface
incorporates a positive heat flux during the night if the URBAN option is chosen in the input file.
Increasing the Bowen ratio to 2 improves the comparison for heat flux, as seen in Figure 5.11,
but the surface friction velocity is still underestimated.
AERMET outputs for Long Beach were compared with measurements made during a field study
conducted in Wilmington in 2005. The measurements were made at Los Angeles County
Sanitation District’s Joint Water Pollution Control Plant (JWPCP) during the period June 16th
and June 30th 2005. The top panel of Figure 5.12 indicates that when AERMET estimates of
heat flux are small or even negative the corresponding observations are positive and large. This
58
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ suggests that AERMET might not be capturing urban heat island effects seen in the observed
heat fluxes. However, AERMET performs well in estimating the distribution of observed surface
meteorological variables when the observed heat fluxes are positive.
These comparisons of AERMET estimates for Riverside and Long Beach with measurements
made at diverse sites in Riverside and Wilmington indicate that the discrepancies between the
measured values and AERMET estimates are generally within a factor of two of each other; such
deviations do not always translate into similar differences in the distributions of the surface
variables. The consequences of such differences need to be evaluated by running AERMOD with
differing surface files.
It is clear that AERMET files generated using data at one site will differ from those generated
using data from another nearby site. Also, AERMET estimates of surface variables such as heat
flux will differ from observations made at the same site, as seen in Figure 5.13. Here AERMET
was run using the observed wind speeds at the upwind site in Riverside. The Bowen ratio was
taken to be unity, and the roughness length was adjusted to provide the best fit between the
estimates and observed values of surface friction velocity. The cloud cover was taken to be 0.5
in the absence of observations. Even under these ideal conditions, the deviation between model
estimates and observations is a factor of two. Note that observed values of the standard deviation
of horizontal velocity fluctuations, σv, are rarely less than 0.5 m/s when the corresponding
estimates are much smaller.
We see in Figure 5.14 similar deviations between model estimates and observations at the central
urban (CU) site in Riverside where Bowen ratio is taken to be 0.5 to account for the possibility
of high latent heat fluxes.
The results presented here indicate the need to conduct sensitivity studies with AERMOD to
understand the impact on design concentrations of unavoidable differences between AERMET
estimates of meteorological variables and corresponding observations at the site of interest.
59
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Table 5.1: Average daytime boundary layer heights over hours 1100 – 1500 LST for each
sub-area.
Sub-Area Average PBL height (m)
ANAH 866AZUS 837BNAP 927BURK 852CELA 861CRES 834CSTA 775ELSI 837FONT 857INDI 985LAHB 859LAXH 777LGBH 774LYNN 774MSVJ 864PERI 871PICO 839POMA 855PLSP 959RDLD 863RESE 849RIVR 861SCLR 822SNBO 866UPLA 860WSLA 768
60
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 5.1: Map of South Coast Air Quality District area with wind roses for AQMD monitoring station data used
to produce AERMOD input files overlaid. Wind roses overlaid on map are also shown in Appendix A.
61
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0
10
20
30
40
50
60
70
ANAHAZUS
BNAPBURK
CELACRES
CSTAELSI
FONTIN
DLAHB
LAXHLGBH
LYNNMSVJ
PERIPIC
OPOMA
PSLPRDLD
RESERIVR
SCLRSNBO
UPLAWSLA
% o
f Hou
rs
< 1 m/s < 0.5 m/s
Figure 5.2: Percentage of winds less than 1 m/s and 0.5 m/s in the AQMD monitoring station wind data used to
produce the *.sfc and *.pfl files.
62
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 5.3: Average daytime values of surface friction velocity over the hours 0900 – 1500 LST for all sub-areas.
63
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 5.4: Average values of free convective scaling velocity over the hours 0900 – 1500 LST for all sub-areas.
64
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 5.5: Average nighttime values of surface friction velocity over the hours 2100 – 0500 LST for all sub-areas.
65
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 200 400-100
0
100
200
300
400
500
AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
AERMET u* (m/s)
Obs
erve
d u *
0 200 400-100
0
100
200
300
400
500
Ranked AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
Ranked AERMET u* (m/s)
Obs
erve
d u *
Figure 5.6: Comparison of measurements made at the at the US site with estimates from AERMET. The top panel compares observations paired in time, while bottom panels compare ranked observations and estimates.
66
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 5 10 15 20 25-0.1
0
0.1
0.2
0.3
Time of day
Hea
t flu
x in
K.m
/s
4/26
AERMETObserved
0 5 10 15 20 250
0.2
0.4
0.6
0.8
Time of day
u * in m
/s
0 5 10 15 20 250
1
2
3
4
Time of day
Win
d sp
eed
in m
/s
0 5 10 15 20 250
100
200
300
400
Time of day
Win
d di
rect
ion
in d
egre
es
Figure 5.7: Comparison of diurnal variation of measurements made at the at the US site with corresponding estimates from AERMET for 4/26/2007.
67
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 5 10 15 20 25-0.1
0
0.1
0.2
0.3
Time of day
Hea
t flu
x in
K.m
/s
3/31
AERMETObserved
0 5 10 15 20 250
0.1
0.2
0.3
0.4
0.5
Time of day
u * in m
/s
0 5 10 15 20 250
1
2
3
4
Time of day
Win
d sp
eed
in m
/s
0 5 10 15 20 250
100
200
300
400
Time of day
Win
d di
rect
ion
in d
egre
es
Figure 5.8: Comparison of diurnal variation of measurements made at the at the US site with corresponding estimates from AERMET for 3/31/2007.
68
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 200 400-100
0
100
200
300
400
500
AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
AERMET u* (m/s)
Obs
erve
d u *
0 200 400-100
0
100
200
300
400
500
Ranked AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
Ranked AERMET u* (m/s)
Obs
erve
d u *
Figure 5.9: Comparison of measurements made at the CU site with estimates from AERMET. The top panel compares observations paired in time, while bottom panels compare ranked observations and estimates.
69
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 200 400-100
0
100
200
300
400
500
AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
AERMET u* (m/s)
Obs
erve
d u *
0 200 400-100
0
100
200
300
400
500
Ranked AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
Ranked AERMET u* (m/s)
Obs
erve
d u *
Figure 5.10: Comparison of measurements made at the at the DS site with estimates from AERMET. The top panel compares observations paired in time, while bottom panels compare ranked observations and estimates.
70
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 200 400-100
0
100
200
300
400
500
AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
AERMET u* (m/s)
Obs
erve
d u *
0 200 400-100
0
100
200
300
400
500
Ranked AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
Ranked AERMET u* (m/s)
Obs
erve
d u *
Figure 5.11: Comparison of measurements made at the at the DS site with estimates from AERMET. Bowen ratio is taken to be 2 instead of the nominal value of 1. The top panel compares observations paired in time, while bottom
panels compare ranked observations and estimates.
71
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
0 200 400-100
0
100
200
300
400
500
AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
AERMET u* (m/s)
Obs
erve
d u *
0 200 400-100
0
100
200
300
400
500
Ranked AERMET heat flux (W/m2)
Obs
erve
d he
at fl
ux
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
Ranked AERMET u* (m/s)
Obs
erve
d u *
Figure 5.12: Comparison of measurements made at Wilmington in 2005 with estimates from AERMET for Long Beach. The top panel compares observations paired in time, while bottom panels compare ranked observations and
estimates.
72
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
101 102 103101
102
103
Estimated heat flux in W/m2
Obs
erve
d he
at fl
ux
10-1 10010-1
100
Estimated u* in m/s
Obs
erve
d u *
10-1 100 10110-1
100
101
Estimated σw in m/s
Obs
erve
d σ
w
10-1 100 10110-1
100
101
Estimated σv in m/s
Obs
erve
d σ
v
Figure 5.13: Comparison of measurements made at the at the US site with estimates from AERMET using wind speeds and temperatures at the site.
73
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
Figure 5.14: Comparison of measurements made at the CU site with estimates from AERMET using wind speeds and temperatures at the site.
74
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________
6 Conclusion
Summary
AERMOD surface (*.sfc) and profile (*.pfl) meteorological input files have been developed for
various sub-areas across the South Coast Air Quality Management District air basin. AERMET,
developed by the U.S. EPA, was used to develop these files. The preceding sections of this report
describe in detail the contents of the *.sfc and *.pfl files (Section 2), the AERMET processor that
creates them (Section 3), the methodology used to process the raw data to create the
meteorological input files for AERMET (Section 4), and the resulting mean wind and
micrometeorological turbulence fields produced by AERMET for each sub-area over the air
basin (Section 5). A comparison of the AERMET results for the Riverside and Long Beach sub-
areas with observations in Riverside and Wilmington, respectively, is also presented (Section 5).
Three post-processors were developed to allow the user to check and modify the *.sfc and *.pfl
files produced by AERMET. Application of these post-processors can allow modification of
these files to partially account for meteorological effects not accounted for in AERMET. The
post-processors are described in detail in Section 4 and Appendix C.
A ‘User’s Guide’ for the full procedure of processing raw meteorological data to produce input
files for AERMET is contained in Appendix C. The User’s Guide also gives guidance on running
AERMET and provides a description of the three post-processors mentioned above to modify the
AERMET output.
All deliverables – the *.sfc and *.pfl files, raw meteorological data, AERMET input files, post-
processing codes, and other miscellaneous files – are provided on the accompanying CD.
Appendix D describes the directory structure of this CD.
75
EC-07-015 – 09-04-17 (I) Final Report (Vol. I) _____________________________________________________________________________________________ Suggestions for Future Work
After carrying out this work and reviewing the results contained in the provided *.sfc and *.pfl
files, we have the following suggestions for future work:
1. The wind data for several of the AQMD stations should be reviewed to check that the
data are realistic. At several stations, for example CSTA, LAHB, LYNN, RESE and
POMO, a large percentage of hours (~ 50%) have winds less than 1 m/s. See discussions
in Section 4 (specifically Table 4-4 and associated discussion) and Section 5 (specifically
Figure 5.2 and associated discussion) for further details.
2. While the hourly values in the AERMET output compared with observations at Riverside
and Wilmington are generally within a factor of two, considerable variability exists in the
comparison, both in space and time. Such variability appears to be inherent to
observations made in urban areas. The impact of this variability on design concentrations
should be investigated by running AERMOD with input files that reflect this variability.
3. The daytime boundary layer depths produced by AERMET for the coastal sub-areas (for
example, WSLA, LAXH, LGBH and CSTA) are probably on average too high because
AERMET does not account for coastal internal boundary layers that result from the
prevailing onshore flow in these sub-areas or for the limiting effects of large scale
subsidence that is prevalent in the South Coast Air Basin. The post-processors produced
in this project can potentially provide suitable corrections to the daytime boundary layer
depths and convective velocity scales in the chosen sub-areas.
4. The sensitivity of design concentrations to reasonable and anticipated variations in such
meteorological variables such boundary layer depth, Bowen Ratio, and different choices
of meteorological station providing wind data for a given sub-area should be investigated
to see how important these factors are in affecting design concentrations. Such an effort
will help identify the most important improvements to be made to the *.sfc and *.pfl files
provided thus far.
76